key: cord-0806985-gksid5rl
authors: Cazorla, María; Herrera, Edgar; Palomeque, Emilia; Saud, Nicolás
title: What the COVID-19 lockdown revealed about photochemistry and ozone production in Quito, Ecuador
date: 2020-08-25
journal: Atmos Pollut Res
DOI: 10.1016/j.apr.2020.08.028
sha: 86d29b1a5ed40fb5ec84268df853e0be8d8b4f1c
doc_id: 806985
cord_uid: gksid5rl

The COVID-19 lockdown presented a peculiar opportunity to study a shift in the photochemical regime of ozone production in Quito (Ecuador) before and after mobility restrictions. Primary precursors such as NO and CO dropped dramatically as early as 13 March 2020, due to school closures, but ambient ozone did not change. In this work we use a chemical box model in order to estimate regimes of ozone production before and after the lockdown. We constrain the model with observations in Quito (ozone, NO(x), CO, and meteorology) and with estimations of traffic-associated VOCs that are tightly linked to CO. To this end, we use the closest observational data of VOC/CO ratios at an urban area that shares with Quito conditions of high altitude and is located in the tropics, namely Mexico City. A shift in the chemical regime after mobility restrictions was evaluated in light of the magnitude of radical losses to nitric acid and to hydrogen peroxide. With reduced NO(x) in the morning rush hour (lockdown conditions), ozone production rates at 08:30–10:30 increased from 4.2 to 17 to 9.7–23 ppbv h(−1), respectively. To test further the observed shift in chemical regime, ozone production was recalculated with post-lockdown NO(x) levels, but setting VOCs to pre-lockdown conditions. This change tripled ozone production rates in the mid-morning and stayed higher throughout the day. In light of these findings, practical scenarios that present the potential for ozone accumulation in the ambient air are discussed.

19 in Ecuador occurred on 29 February 2020 in the coastside province of Guayas. As the infection 51 began to propagate within the country, the government issued school closures at the national level 52 on 13 March 2020. On 17 March, a decree was issued to restrict citizens' free mobility through a 53 21:00-05:00 national curfew. In Quito, the local government suspended public transportation 54 (mainly composed by a fleet of diesel buses) on 17 March. The government further hardened 55 mobility restrictions on 25 March by imposing a 14:00-05:00 curfew in the entire country. 56

Circulation of private vehicles was limited by the last digit of their plate number. This restriction 57 continued into April, although delivery vehicles were allowed to drive until 19:00. These measures 58

had an impact on the levels of NO x and VOCs and on the chemistry of ozone production, which we 59 discuss in this paper. 60 61 Ozone production has been studied comprehensively by many authors (i.e., Haagen-Smit et al., 62

1956; Finlayson-Pitts and Pitts, 1977; Thornton et al., 2002) . Briefly, during daylight hours, NO 2 63 photolyzes and leads to the formation of ozone and NO in a 1:1 stoichiometric proportion. 64

Subsequently, ozone is titrated by NO. The latter two reactions (summarized in Rx. 1) yield a null 65 cycle in the clean background atmosphere. In the urban atmosphere, VOCs are oxidized by the 66 hydroxyl radical (OH). As a result, hydroperoxy (HO 2 ) and organic peroxy radicals (RO 2 ) are 67

formed and react with NO to produce NO 2 . These reactions produce secondary organic radicals 68 (RO'), which continue to oxidize towards additional production of NO 2 (propagation chain 69 compactly represented by Rx. 2 to 4). This NO 2 , that forms outside of the titration step of ozone 70

with NO, rapidly undergoes photolysis and is converted into ozone. Ozone production continues to be fueled in a catalytic fashion by rapid cycling of OH into HO 2 , 86 until a termination step takes place. Termination can occur due several mechanisms of the type 87 NO x -HO x or HO x -HO x . Two of the most studied reactions (Sillman, 1995) are formation of nitric 88 acid and hydrogen peroxide represented by Rx. 5 and 6, respectively (M in Rx. 5 is the altitude 89 dependent abundance of background air molecules). The rate of radical losses due to these 90 mechanisms can be quantified through Eq. 2 and 3 and their magnitudes can be used to assess if the 91 ozone production regime is NO x -saturated (Rx. 5 dominates) or NO x -limited (Rx. 6 dominates) 92 (Kleinman et al., 2001; Kleinman, 2005 The organization of the paper is as follows: in the methods section, we describe the measuring site 121 as well as air quality and meteorology data sets. Furthermore, the strategy used to estimate VOCs 122 and photolysis frequencies is described in detail. In addition, we describe model details and 123

constraints. In the results section, we discuss dependencies encountered among radical abundances, 124 NO x levels and ozone production rates, before and after the COVID-19 lockdown. We include a 125 practical view of scenarios under which increased rates of ozone production could lead to high 126 ozone days in Quito. In the conclusions section, we present the main lessons revealed by such 127 dramatic change in precursors in regard to the production of ozone and its overall effect on air 128 quality. 

We present 10-minute data for NO, NO 2 , and ozone from 1 January to 30 April 2020. Details 146 relative to air quality instrumentation are described in previous work (Cazorla, 2016) . Briefly, NO x 147 measurements were taken with a Teledyne T200 instrument. The instrument is periodically 148 calibrated at EMA using mixtures prepared with a certified NO standard and zero air. Measurements 149 are also intercompared against those of a neighboring City station. From calibrations, 1-σ 150 uncertainty in measurements is 5%. The low detection limit of this instrument is 0.4 ppbv. Thus, 151 only data greater than 0.4 ppbv were used. Ozone is continuously measured with a Thermo 49i 152

analyzer. Ozone measurements are periodically checked by intercomparing against ground station 153 measurements taken with conditioned electrochemical concentration cells used at EMA for vertical 154

profiling (Cazorla, 2017) . Beginning February 2020, ozone measurements are also checked 155

periodically against a new 2B Technologies sensor model 205. The limit of detection of the 49i 156 sensor is 0.5 ppbv. From intercomparisons, 1-σ measurement uncertainty is 8%. In addition to 10-157 minute time series, we present ozone, NO and NO 2 mean diurnal variations (MDV) prepared by 158 overlapping 10-minute data each month and obtaining an average every hour. 159 160

Physical meteorology measurements (10-minute data) taken at EMA from 1 January to 30 April 161 2020 (temperature, pressure, relative humidity, solar radiation, and precipitation) are also presented. 162

Details of the meteorological context within which this work was developed are included in the 163

Supplementary Material (SM), Appendix S1 ( with CO to obtain VOC/CO ratios for: propane, 3-methylpentane, n-butane, n-hexane, ethene, 199

propene, benzene, toluene, m-xylene, and ethylbenzene. Ratios range from 0.0062 (3-200 methylpentane) to 0.06 (propane). All factors and data are presented in the SM (Appendix S3, Table  201 S3). Subsequently, we used Quito CO observations to derive VOCs and constrain the chemical box 202 model, as discussed in section 2.4. Appendix S4, Fig. S4 ). This factor was applied to the solar radiation time series taken at EMA in 215 order to estimate J NO2 . These calculations yielded upper and lower limits for J NO2 (Trebs' method  216 and Mexico City measurements, respectively). Hence, we present the average of both methods as 217 J NO2 estimates and use the difference between the boundaries as a measure of 1-σ uncertainty (+/-218 17% at noon and +/-32% in the mid-morning and afternoon). Subsequently, a simple scaling factor 219 of 4.67x10 -3 (obtained from simple observation) that relates J NO2 to J O3->O1D was extracted from the 220 work by Li in order to account for the effect of clouds. 239 240

The model was run from 1 January to 30 April 2020 in 10-minute time steps. Model runs were 241 performed at EMA in a system that has a Core i7 processor with 32 GB of RAM memory and a 242 MATLAB license owned by USFQ. The following quantities from the model output were used in 243 analyses: radical abundance (OH, HO 2 , and RO 2 ), concentrations of secondary species 244 J o u r n a l P r e -p r o o f (formaldehyde and HONO), corresponding reaction coefficients, and frequencies of photolysis. A 245 complete description of model output can be found elsewhere (Wolfe, 2020) . 246 247 2.5 Radical and ozone production rates 248 249

The chemical production of HO x radicals, p(HO x ), was obtained considering the main mechanisms 250

for the production of OH and HO 2 , namely ozone photolysis followed by reaction Ozone production rates, calculated with the above procedure, were also determined for the 271 hypothetical case that NO x levels stayed low, while CO and VOCs were those of normal pre-272 lockdown conditions. This estimation was done to discuss more extensively potential scenarios that 273 could lead to a more permanent shift in the chemical regime of ozone production in Quito. This test 274

was done with NO x , ozone, and meteorological data from 14 March to April, but mimicking CO and 275

VOCs levels from before the quarantine. presented in Fig. 2a and b (the SM, Appendix S5, Fig. S5 contains time series). 290 291

As presented, morning rush hour levels of NO approach 100 ppbv (10-minute data). The local 292 legislation does not impose NO x emission controls on vehicle exhausts, which explains rush hour 293 NO levels of the magnitudes depicted in Fig. 2 . In addition, boundary layer depths during morning 294 rush hours are shallow (below 500 m from 07:00-09:00), as calculated from station measurements 295 (Cazorla and Juncosa, 2018) (SM, Appendix S6, Fig. S6 ). This physical factor contributes to 296 concentrating primary emissions near the ground in the morning. Under quarantine conditions, NO 297 maxima plummeted by a factor of five from mid-March into April. In mid-March the reduction was 298

greater, but in April permits were implemented for delivery vehicles and medical emergencies, 299

which somewhat increased the traffic flow. 300 301

In regard to carbon monoxide (Fig. 3) , average changes after the initiation of the lockdown period 302 are evident as levels dropped by a factor of 1.86 (from about 973 to 523.5 ppbv in the morning rush 303 hour). It is important to remark that CO was not measured at EMA station, which is a limitation. 304

However, CO levels are similar across different stations managed by the local network as presented 305

in their annual reports (Secretaría de Ambiente, 2018). Therefore, it is reasonable to assume that an 306 average within Quito yields a good estimate of mean CO levels before and after the COVID-19 307 lockdown. Figure 3 also contains average levels (before and after the lockdown) of propene, one of 308 the ten VOCs derived from VOC/CO ratios. Fig. 3 shows that propene dropped proportionally along with CO. The SM (Fig  315  S3 ) contains data of all VOCs derived for this work. 316 317

While primary precursors decreased considerably after 13 March, the day school closures were 318 issued, and stayed low into April, ozone levels did not experience evident changes. Fig. 4 shows 319 that 10-minute maxima peaked at 45-50 ppbv from January to April 2020. From diurnal variations, 320 average peak ozone before and after 13 March was about 33 ppbv. Daily data (SM, Appendix 7, 321 Fig. S7 .1) shows that noontime maxima in January were lower, on average, than levels in February  322 and March, even though it was the sunniest month of the trimester. Also, titration with NO at 323 around 06:00 (local time) was reduced during the quarantine due to reduced morning rush hour NO. 324

A more dramatic comparison is average ozone in January and April, which are comparable (SM, 325 Fig. S7 .2), even though April was considerably more cloudy and had more precipitation. Thus, 326

questions arise in regard to levels of ozone from January to April in connection with the drastic 327 reduction in emission precursors after 13 March. An explanation based on photochemistry follows. 328 329

Photolysis reactions in the atmosphere directly depend on the amount of solar radiation available for 332 the dissociation of molecules (Seinfeld and Pandis, 2006) . In particular, ozone formation is highly 333 sensitive to NO 2 photolysis. Likewise, HO x radicals are formed due to photolysis of ozone and other 334 important species in the urban atmosphere, such as HONO and formaldehyde. In the present case, 335 the pandemic lockdown took place during months that seasonally transitioned from sunny skies in 336

January into more cloudy conditions with increased precipitation in April. This changing 337 meteorological conditions resulted in variability of J NO2 diurnal profiles across the months, before 338

and after the lockdown, as depicted in Fig. 5a . Thus, before the quarantine, January was consistently 339 sunny, which led to J NO2 maxima at noon, while the first half of March had sunny mornings but 340 cloudy afternoons. The second half of March was comparable to February in terms of noontime 341 J NO2 , but April was mostly cloudy. In order to compare photochemical quantities before and after 342 the lockdown in a way that the effect of precursors can be assessed, while keeping physical factors 343 constant, we used data that corresponds to sunny conditions in the entire time series (the complete 344 J o u r n a l P r e -p r o o f data set was treated equally, details of data filtering in the SM, Appendix 8, Fig. S8 ). Hence, mean 345 J NO2 profiles from January to 13 March and from 14 March to April, that correspond to sunny 346 conditions, are presented in Fig. 5b . Therefore, radical abundance and ozone production rates are 347 discussed in terms of equal conditions of J NO2 . 348 349

3.3 Radical abundance and production 350 351

Average OH radical abundance under sunny conditions in the periods before and after the COVID-352

19 lockdown are depicted as diurnal profiles in Fig. 6a . Under post-lockdown conditions (14 March  353 to April), OH radicals began to rise as early as 07h30 and remained higher than under regular traffic 354

conditions throughout the day. In the morning, between 08:30-11:30, OH ranged from 0.07-0.55 355 pptv before 14 March, while afterwards it rose to 0.23-0.68 pptv. This feature is not due to physical 356 reasons because solar radiation conditions are kept identical. Hence, additional radical abundance is 357 due to atmospheric chemistry as the chemical depletion of radicals due to high NO x levels 358

decreased during the lockdown, in particular in the morning rush hour (this feature is explained 359

further below with magnitudes of radical loss). Likewise, the abundance of hydroperoxy and 360 organic peroxy radicals (HO 2 and RO 2 ) increased ( Fig. 6b and c) and OH was faster during the lockdown throughout the day as given by HO 2 /OH ratios in Fig. 7a . 365

The test simulation with pre-lockdown levels of VOCs and post-lockdown (reduced) levels of NO x 366

shows that in the absence of elevated NO x , OH is comparable to observed post-lockdown 367

conditions, but at midday cycling between HO 2 and OH has the potential to almost triple with 368 regard to regular traffic conditions (Fig. 7a) . For this reason, in the test simulation noontime 369 abundance of OH is lower, but HO 2 and RO 2 abundances are substantially higher. (including the reduced-NO x test) (Fig. 8a to c) . 378 379

Formaldehyde photolysis is a major source of hydroperoxy radicals in the urban atmosphere 380

(complete reactions can be found in Seinfeld and Pandis, 2006) . Formaldehyde is formed along the 381 oxidation chain, when VOCs react with OH. In the present case, we used the formaldehyde model 382 output (SM, Appendix 9, Fig. S9 ) and its corresponding model J HCHO in order to calculate this 383 contribution to p(HO x ) (second term of Eq. 4, depicted by the red dotted line in all panels in Fig. 8 ). 384

During the study time period, formaldehyde photolysis was more important after the lockdown, 385 mainly in the morning. Thus, magnitudes were 1.8 and 1.6 times higher at 09:00 and 10:00, 386 respectively (the SM, Appendix 10, Fig S10 shows curves in Fig. 8 overlapped by contribution) . At 387 noon, this source of radicals was about 0.3 pptv s -1 throughout the study time period as NO before 388 the lockdown decreased after the morning rush hour, while during the lockdown stayed low (Fig. 2) . 389

The test simulation with pre-lockdown levels of VOCs, but reduced (post-lockdown) levels of NO x 390 (Fig. 8c) further backs this chemical result as the photolysis of formaldehyde became as important 391

as ozone photolysis in regard to producing HO x radicals. These results are consistent with previous 392 work that shows how in urban environments (for example in Mexico City) the contribution of 393 formaldehyde photolysis accounts for up to 40% of HO x radicals (Shirley et al., 2006) . In the 394 present case, we estimate that this source contributes with about 34% to the total radical production 395 rate. 396 397

HONO photolysis is also a significant source of radicals, mainly in the morning during the traffic 398 rush hour, as demonstrated by former studies (Dusanter et al. 2009 ; Ren et al., 2013) . As in the 399 previous case, this contribution was calculated with the third term of Eq. 4 and using the MCM 400 model output for HONO (SM, Fig. S9 ) and J HONO . This contribution was 1.4 times higher before 401 the lockdown as rush-hour NO x was five times higher. The test done with reduced levels of NO x is 402 consistent, as this contribution was comparable to post-lockdown conditions. 403 404

In this study we rely in the robustness of observation-based methods to find estimates of radical 405 production rates from the first source (O 1 D+H 2 O), while the other two sources come from 406 modeling. Hence, from propagation of error, we estimate that uncertainty (1-σ) in the first source is 407 32% (considering uncertainties in J O3->O1D as well as in ozone and water vapor measurements 408 reported in the methods section). For radical production due to formaldehyde and HONO photolysis 409

(as well as for radical abundance), at the moment we assume model uncertainty reported in the 410 literature (i.e., Brune et al., 2019) of 35%. Likewise, we estimate that uncertainty in other 411 photochemical quantities (radical abundance and ozone production) is similar, but in the future 412 these estimations need to be re-checked as measurements of physical and chemical quantities 413 become available in the study area. 414 415

3.4 Ozone production rates and regimes 416 417

The mean diurnal variation of ozone production rates before and after the lockdown are depicted in 418 Fig. 7b . During the quarantine, ozone production rates at 08:30-10:30 increased from previous 4.2-419 17 to 9.7-23 ppbv h -1 . These results are consistent with the former section in the sense that under 420 regular traffic conditions high morning NO x levels constrain ozone production due to a lack of 421 sufficient radicals to react with NO and produce ozone. Quarantine conditions induced a chemical 422

shift that led to increased availability of HO 2 and RO 2 radicals that combined with reduced morning 423 NO x led to higher ozone production rates. Radical abundance being at the core of the observed shift 424 in regimes of ozone production can be assessed by looking into the magnitude of paths for radical 425 losses to nitric acid (L1) and to hydrogen peroxide (L2). Fig. 9 depicts the ratio of L1 to the total 426 loss L1+L2. When the radical loss to hydrogen peroxide is as important as the radical loss to nitric 427 acid, the ratio becomes 0.5. A ratio much greater than 0.5 has been shown to indicate a VOC-limited 428

(NO x -saturated) regime (Kleinman et al., 2001; Kleinman, 2005; Ren et. al., 2013) . As presented, 429

L2 was more significant under quarantine conditions, while this term had little effect from January 430 to 13 March. Consequently, the time period from January to the first half of March was strongly 431 NO x -saturated. In contrast, after 13 March, losses to hydrogen peroxide were of increasing 432

importance, which confirms a shift in the chemical regime of ozone production. The test calculation 433 done with reduced NO x at post-lockdown levels, but mimicking pre-lockdown VOCs, pushed even 434 further this shift towards the NO x -limited zone as ozone production tripled in the mid-morning (Fig. 435 7b) and stayed high, while L1/(L1+L2) became even lower (Fig. 9) . The SM (Appendix 11, Fig.  436 S11) shows individual magnitudes of L1 and L2 for all cases. 437 438

The dependence of p(O 3 ) with NO, before and after the lockdown, as well as for the reduced NO x 439 scenario, is depicted in Fig. 10 with indication of p(HO x ) magnitudes. As presented, under regular 440 conditions ozone production is strongly suppressed under a NO X -saturated regime (Fig. 10a) . In this 441 regime, p(O 3 ) decreases with increasing NO, whose levels approach 100 ppbv. After 13 March (Fig. 442 10b), the drop in NO emissions augmented the rate of ozone production, which steadily grows at 443 low NO and at the higher p(HO x ) range, and then decreases as NO continues to increase. With 444 reduced NO, but "normal" VOCs, this same dependence occurs (Fig. 10c) (Fig. 10b  448 and c). This final analysis is consistent with seminal work on ozone production (Thornton et. al., 449 2012) that demonstrates the high non-linearity of ozone production rates with NO when data is 450 sorted by p(HO x ) levels. A second important lesson is that a reversal in emission precursors in Quito is capable of shifting 469 the chemical regime to the NO x -limited zone, which results in higher ozone production rates. 470

Quarantine conditions, a current reality and one that could repeat itself in the near future, supply 471 higher ozone production rates. Average morning magnitudes, that doubled those before the 472 quarantine, were capable of sustaining the same levels of ozone in Quito, even in months that are 473 seasonally cloudier and with more abundant precipitation. Thus, meteorological conditions played a 474 role in the ozone mass balance providing less sunny days and cleaning the atmosphere through wet 475 deposition. However, if confinement orders took effect in the hot and dry summer months of August 476

and September, higher ozone production rates at low NO x levels could contribute to the 477 accumulation of ozone in the boundary layer, in particular under conditions of stagnant air. This is a 478 realistic scenario that needs to be watched closely under the current reality of a pandemic that poses 479 the need of periodical quarantines to control the spread of the disease. 480 481

Thirdly, results obtained from calculations with reversed levels of precursors shed light to 482

investigating the chemical reasons that underlie few observed episodes of high ozone in Quito. 483 These events have been associated to wildfires (for example on 1 October 2018), as it was recently 484

presented in a preliminary study (Cadena et al., 2019) . However, the true chemical nature of such 485

events needs yet to be scrutinized through studies that involve measurements and modeling. 486

Although the complex effect of advection of air masses from biomass burning regions is beyond the 487 scope of this work, it is important to document that the current study advances the topic of potential 488 scenarios that cause a shift in the chemical regime of ozone production due to drastic changes in the 489

proportion of organic compounds with respect to NO x . 490 491

Finally, a situation that could potentially cause increased ozone is the hypothetical case that NO x 492 emission controls in vehicles were to be imposed. Such scenario would cause a shift of the sort 493 mimicked by the reduced-NO x calculation, in which normal levels or organic compounds combined 494 with permanently low NO x lead to higher ozone production rates. This scenario emphasizes on the 495 fact that environmental practices, that have to do with emission controls, need to carefully take into 496 account that the regime of ozone production depends on the local makeup of pollutants in the 497 ambient air and is non-linear with respect to precursor levels. 498 499

Even though intensive work that incorporates measurement campaigns to constrain chemical and 500 transport models still needs to be developed, the present contribution points out for the first time to 501 specific conditions and identifies practical scenarios under which the chemical regime shifts 502 towards higher rates of ozone production. 503 504

4. Conclusions 505 506

The COVID-19 mobility restrictions and lockdown, that initiated with school closures on 13 March 507

in Quito (Ecuador), marked a shift in primary emissions that was used to reveal the chemical nature 508 of ozone production under regular conditions as well as under reduced levels of precursors. First, 509 low ozone production rates, in spite of abundant urban emissions and equatorial solar radiation, are 510 the rule as radicals are quickly depleted by loss mechanisms of the type NO x -HO x . Results indicate 511

that, under normal traffic conditions, radical loss to nitric acid dominates, in particular in the 512 morning rush hour, when NO spikes (10-minute data) approach 100 ppbv. In contrast, post-513

lockdown NO x levels decreased by a factor of five during daytime. This shift led to less radical 514 losses to nitric acid, while the loss to hydrogen peroxide became increasingly important. Thus, the 515 abundance of HO 2 and RO 2 increased from a total of 4.2 pptv in the mid-morning before the 516 lockdown to 16.1 pptv during the quarantine, while at noon it increased from previous 23.6 to 39 517 pptv. Consistently, OH in the morning at 08:30-10:30 increased from 0.07-0.37 pptv before the 518 lockdown to 0.23-0.61 pptv afterwards, while p(O 3 ) increased from 4.2-17 to 9.7-23 ppbv h -1 , 519

respectively. From observations, magnitudes of pre-lockdown ozone production rates factored with 520 dilution within the boundary layer and advection explain generally low ambient ozone in Quito, but 521 further work needs to be developed to better understand transport effects. As per ozone production 522 during the lockdown, there was seasonally more cloudiness and precipitation during mid-March and 523

April, which helped clean the atmosphere through wet deposition. Hence, meteorological conditions 524 were favorable at preventing ozone accumulation during this period. However, if a quarantine were 525 to take effect during the warmer summer months of August to mid-September, especially if 526 conditions of stagnant air and temporary drought develop, increased rates of ozone production 527 would pose a threat of accumulation in the ambient air. To test further the effect of a shift in 528 emissions that could lead to even higher ozone production rates, a simulation with post-lockdown 529 NO x levels, but pre-lockdown VOCs and sunny conditions yielded a total of 52 pptv of HO 2 and 530 RO 2 at noon and p(O 3 ) of 33 ppbv h -1 , on average (higher than 40 ppbv h -1 as 10-minute data). A 531 scenario that would cause a permanent shift towards this regime would be if NO x emission controls 532 on vehicle exhausts, currently not applied, were to be enforced. Although such scenario is not a 533 current risk, from a scientific and academic standpoint it is prudent to remark that such situation 534 would have an impact on air quality as it would cause a shift in the chemical regime towards 535

producing ozone at higher rates. In spite of many limitations, the change in emissions due to the 536 COVID-19 lockdown was substantial enough to reveal critical information in regard of precursor 537 levels that cause a shift in the regime of ozone production. This contribution advances our 538 understanding of the underlying chemistry of photochemical smog in Quito, Ecuador. 539 540

Author Contributions 541 542 J o u r n a l P r e -p r o o f

Ozone structure over the equatorial Andes from balloon-borne observations and 594 zonal connection with two tropical sea level sites

Planetary boundary layer evolution over an equatorial Andean valley: 598 A simplified model based on balloon-borne and surface measurements

Measurements of OH and HO 2 604 concentrations during the MCMA-2006 field campaign -Part 2: Model comparison and radical 605 budget

Coronavirus lockdown leading to drop in pollution across Europe

The Emission Characteristics of a Diesel Engine During 613 Start-Up Process at Different Altitudes. Energies

A meteorological overview of the MILAGRO field campaigns

The chemical basis of air quality: Kinetics and mechanism 622 of photochemical air pollution and application to control strategies

Coronavirus disease 627 2019: International public health considerations

Ozone formation in photochemical oxidation 631 of organic substances

Air 634 pollution, weather, and associated risk factors related to asthma prevalence and attack rate

Proyecciones referenciales de población a 638 nivel cantonal-provincial

Chemistry of HO X radicals in the upper 642 troposphere

Non-methane hydrocarbons in the atmosphere of Mexico City: Results of the 646 2012 ozone-season campaign

The tropospheric degradation of volatile organic 650 compounds: a protocol for mechanism development

Sensitivity of ozone production rate to ozone precursors

The dependence of tropospheric ozone production rate on ozone precursors

Normal atmosphere: Large radical and formaldehyde concentrations predicted

Aerosol effects on the photochemistry in Mexico City 664 during MCMA-2006/MILAGRO campaign

A Historical Overview of the Ozone Exposure Problem

Airborne Nitrogen Dioxide Plummets Over China

Impact of altitude on emission rates of ozone 675 precursors from gasoline-driven light-duty commercial vehicles

La Calidad del aire en Quito se mantiene en niveles óptimos

Atmospheric oxidation chemistry and ozone production: Results 685 from SHARP

Protocol for the development of 689 the Master Chemical Mechanism, MCM v3 (Part A): tropospheric degradation of non-aromatic 690 compounds

Informe de la Calidad del Aire del Distrito Metropolitano Quito 694

Datos Horarios Históricos Red Monitoreo Aire

Atmospheric chemistry and physics: from air pollution to climate 702 change

Informes de Situación e Infografias 705 -COVID 19 -desde el 29 de Febrero del 2020

Mexico City Metropolitan Area (MCMA) during

The use of NO y , H 2 O 2 , and HNO 3 as indicators for ozone-NO x -hydrocarbon 715 sensitivity in urban locations

Ozone 720 production rates as a function of NO x abundances and HO X production rates in the Nashville urban 721 plume

Relationship between the NO 2 photolysis frequency and the solar 725 global irradiance

The Framework for

Atmospheric Modeling (F0AM) v3.1. Geosci. Model Dev

Overview of the Framework for 0-D Atmospheric Modeling (F0AM)

Blue dots mark locations of three stations managed by Quito's air 761 quality network (Cotocollao, Belisario and Los Chillos), whose CO data where averaged. b) Quito 762 location relative to Ecuador

measured at EMA USFQ before (green crosses and blue line) and after (gray dots and 766 red line) the COVID-19 lockdown. b) The same, but for NO 2 . Crosses and dots are 10-minute data 767 and lines are median diurnal variations (MDV). Dates before and after the lockdown were

-minute data), in NO space sorted by magnitudes of 888 p(HO x ) for: a)

Reduced-NO x simulation with post-lockdown NO x and pre-lockdown VOCs